Highly Efficient Zinc-Based Metal–Organic Framework for Adsorptive Removal of Congo Red: Experimental Investigation and AI-Driven Validation

Mitigating hazardous environmental hazards, such as organic dyes present in water, remains a pressing challenge. In this study, a highly efficient zinc-based Metal-Organic Framework (MOF) is successfully employed for the adsorptive removal of Congo Red (CR) from aqueous solution. Experimental design was validated using Machine Learning (ML), and key parameters are optimized to model and predict adsorption performance. A facile synthetic approach is adopted for the synthesis of zinc benzenetricarboxylate (ZnBTC) MOF via a hydroxy double salt (HDS) intermediate under ambient conditions. The resultant ZnBTC MOF consists of zinc nodes interconnected by carboxylate linkers, forming a three-dimensional porous framework with well-defined channels and an appreciable surface area. This structural architecture was confirmed through scanning electron microscopy, transmission electron microscopy, Fourier transform infrared spectroscopy, X-ray diffraction, N₂ adsorption-desorption analysis, and X-ray photoelectron spectroscopy. The material demonstrates remarkable removal efficiency of over 94% across various CR concentrations in a contact time of 60 min. A detailed investigation carried out to evaluate the influence of various key parameters, such as initial concentration of dye, temperature, solution pH, adsorbent dosage, and the influence of coexisting ions, on the dye uptake performance of ZnBTC demonstrates the robustness of this material as a dye adsorbent. To further elucidate and predict the adsorption performance, the experimental data set generated from the investigation is then integrated into ML. Four basic approaches, namely, Extreme Gradient Boosting (XGBoost), Random Forest (RF), Artificial Neural Network (ANN), and Linear Regression (LR), are compared. The XGBoost model delivers the strongest performance with a test <i>R</i>² value of 0.968 and establishes the catalyst dosage and contact time as the governing variables. This establishes the superiority of tree-based ensemble machine learning algorithms to predict MOF-based adsorption performance, further confirming that such approaches are particularly well-suited for small, structured data sets. Thermodynamic, isotherm, and kinetic analyses confirm the spontaneous and predominantly chemisorptive nature of CR uptake, with additional contributions from hydrogen bonding, mass-transfer effects, and π-π interaction effects. Owing to its facile preparation, biocompatibility, and sustained efficiency across varied operational conditions, ZnBTC emerges as a promising and efficient adsorbent for CR removal in wastewater treatment systems. Moreover, the combined experimental and AI-driven approach yields a reliable decision-support tool for adsorption studies, underscoring the significance of this work in bridging materials science with AI for a reliable decision-support framework for wastewater remediation studies.